People Inc. held its Q2 call on Tuesday and gave the industry its first clean look at what a year of AI search does to a large publisher’s P&L. Core sessions, meaning unique visitors to its main brands, fell 22% year over year. Google search traffic fell 40%. Google now sends about 21% of the company’s total traffic, down from 25% just last quarter and from something in the neighborhood of two-thirds not long ago.
Then the part that got almost no pickup: session-based advertising revenue was flat.
Not down 22%. Flat. A fifth of the audience walked out of the building and the money that audience generated stayed where it was. Digital revenue overall grew 6%, the eleventh straight quarter of growth, and digital adjusted EBITDA grew 18% with margin expanding to 26% from 23%. This is a company whose traffic chart looks like a cliff and whose earnings look fine.
Every writeup I saw led with Vogel’s answer on whether People Inc. would block Google’s AI crawlers. Fair question, good quote. But the crawler debate is a negotiating posture, and the flat revenue line is a fact about how advertising is currently being priced. One of those is more useful on Monday.
The missing fifth of the audience got repriced
CFO Tim Quinn said it plainly on the call: “While our sessions are down, our rates are up significantly. Our rates are up as a function of the quality content and the performance of our sales.”
If impressions drop roughly a fifth and revenue holds, the effective rate on what’s left rose about a quarter. People Inc. sold fewer ads for enough more money to cover a 22% hole. That is a supply story, and it’s the first hard evidence I’ve seen that the AI-search decline is doing something to premium editorial CPMs other than crushing them.
It makes sense once you say it out loud. AI answers are absorbing the low-intent, high-volume, one-and-done visit. The person who wanted to know how long to boil an egg is never coming to Allrecipes again, and that visit was never worth much anyway. What’s left on a People or Investopedia page skews toward someone who chose to be there. Fewer impressions, better ones, and a fixed pool of advertisers who still need to reach adults on brand-safe editorial. Supply contracts, demand doesn’t, price moves.
I watched broadcast do a version of this to itself
When I was at Crispin Porter + Bogusky, network television was in the middle of losing its audience to cable, then to DVRs, then to everything else. The ratings for an average primetime hour fell for twenty straight years. The cost to buy that hour did not fall for twenty straight years. Live sports and the Super Bowl kept setting price records the entire time the medium was supposedly dying, because as mass reach got rarer, the remaining pieces of it got more expensive.
Scarcity pays whoever owns the thing buyers ask for by name. It does nothing for everyone else. The same two decades that made a Super Bowl unit priceless wiped out a long tail of syndicated inventory nobody requested specifically. People Inc. owns People, Allrecipes, Investopedia, Southern Living, Byrdie. Those get requested by name. If you run a mid-tier site whose traffic came from ranking for a query, you are getting the identical 40% Google decline with none of the pricing power, and your rate card is not going to save you. Two things can be true: the AI-search reckoning is real, and it is going to hit publishers with brands very differently than publishers with rankings.
Forty-three percent of the money now comes from things you sell by hand
The other half of People Inc.’s answer is a mix shift, and it’s the part that costs an organization the most to execute. Non-session revenue grew 16%, from $108 million to $125 million, and went from 39% of digital revenue to 43%. Vogel attributed the growth to Apple News licensing including AI partnerships, social programs, events, and D/Cipher, the company’s contextual targeting product.
Look at that list as a seller rather than as a reader. Licensing deals, AI content partnerships, event sponsorships, custom social programs. Not one of those clears through an auction. Every one is a named-account, multi-meeting, custom-scoped negotiation with a longer cycle, a bespoke deliverable, and a different buyer than the one holding the programmatic budget. Quinn described packaging session and non-session inventory together across web, social and events, which is an integrated-program sale, the kind agencies staff with strategists.
You cannot get there by retraining the team you have on a Tuesday. It means different hires, a different comp plan, a different pipeline review, and eighteen to twenty-four months of pain while the new motion learns to forecast. People Inc. has been building this since well before the AI story broke, which is the actual reason the quarter looks the way it does. The publishers starting this rebuild now are starting it while the traffic is already gone, and they will be selling integrated programs with a sales org that only knows how to sell impressions.
Two things can be true about a rate increase
I don’t want to oversell this. A scarcity premium holds only while demand holds, and advertising demand is not a fixed quantity sitting politely in place. Budget follows attention eventually, and attention moved to AI answers, retail media and social video. Quinn credited rates to “the performance of our sales,” which is a sales-execution answer, and sales execution is not a moat. Somebody has to renegotiate those rates again in twelve months against another year of session decline, and the argument gets harder each time.
Vogel was honest about the bind on the call. People Inc. would like more bargaining room with the AI companies using its content, and it still can’t block Google’s crawlers, because turning off the AI crawler turns off search along with it. “We are nearly on the other side of search being a material driver of value for us. But we’re not there yet,” he said, adding that the company isn’t “galloping on a high horse trying to make a point” and just wants a fair economic deal for its content. That’s a company describing exactly how much rope it has.
Split volume from rate before your next planning cycle
If you sell media, go find out whether your flat revenue line is flat because volume held or because rate covered a volume loss. Those two situations look identical on a dashboard and require completely different plans. One means you’re fine. The other means you have a pricing runway of unknown length and should be spending it building whatever your version of non-session revenue is.
If you buy media, expect premium editorial CPMs to keep climbing while the impressions available keep shrinking, and start asking sellers what happened to their sessions before you accept the new rate. You may be paying a scarcity premium to reach a smaller number of the same people.
And if you’re running a GEO program, hold this quarter’s numbers next to it. I’ve written that a blended AI-visibility score hides which engine you’re losing and that Alphabet’s own top line will be the last place the shift shows up. Add this one to the pile. Every AI-visibility tool on the market reports traffic, citations and share of answer. None of them report what your remaining traffic is worth per visit. People Inc. just demonstrated that the second number can move enough to make the first one survivable, and nobody is selling you a dashboard for it.
Next quarter I’ll be watching whether Vogel reaches the point where he can turn the crawlers off and mean it. Right now 21% of his traffic says he can’t.